Skip to main content
Internet Archive's 25th Anniversary Logo

Full text of "The Accuracy of Morphological Decomposition of Active Galactic Nucleus Host Galaxies"

See other formats

Accepted for publication in Ap. 

The Accuracy of Morphological Decomposition of AGN Host 


B. D. Simmons and C. M. Urry 

Astronomy Department, Yale University, New Haven, CT USA 

simmonsOastro . yale . edu 


In order to assess the accuracy with which we can determine the morpholo- 
gies of AGN host galaxies, we have simulated more than 50,000 ACS images 
of galaxies with z < 1.25, using image and noise properties appropriate for the 
GOODS survey. We test the effect of central point-source brightness on host 
galaxy parameter recovery with a set of simulated AGN host galaxies made by 
adding point sources to the centers of normal galaxies. We extend this analysis 
and also quantify the recovery of intrinsic morphological parameters of AGN host 
galaxies with a set of fully simulated inactive and AGN host galaxies. 

We can reliably separate good from poor fit results using a combination of 
reasonable error cuts, in the regime where (-^^hostZ-^PS) > 1 : 4. We give quan- 
titative estimates of parameter errors as a function of host-to-point-source ratio. 
In general, we separate host and point-source magnitudes reliably at all redshifts; 
point sources are well recovered more than 90% of the time, although spurious 
detection of central point sources can be as high as 25% for bulge-dominated 
sources. We find a general correlation between Sersic index and intrinsic bulge- 
to-total ratio, such that a host galaxy with Sersic n < 1.5 generally has at least 
80% of its light from a disk component. Likewise, "bulge- dominated" galaxies 
with n > 4 typically derive at least 70% of their total host galaxy light from a 
bulge, but this number can be as low as 55%. Single-component Sersic fits to 
an AGN host galaxy are statistically very reliable to z < 1.25 (for ACS survey 
data like ours). In contrast, two-component fits involving separate bulge and disk 
components tend to over-estimate the bulge fraction by ~ 10%, with uncertainty 
of order 50%. 

Subject headings: methods: data analysis — galaxies: active — galaxies: nuclei 
— galaxies: bulges — galaxies: fundamental parameters 

- 2 - 


Morphological analysis of two-dimensional light profiles of galaxies in large data surveys 

provides detailed informat i on about galaxy populat i ons and their evolu tion (e.q. JSimard et al. 


Ravindranath et al. 


Joeee et al. 


Sheth et al. 


large and small galaxy samples show that two-dimensional para metric and non-parametric 

morphological analysis of no r mal galaxies is extreme l y robust (iMar 

Grahamll200ll ; IConselicell2003l ; IXruiillo &: Aguerrill2004l ; iHaussler et al. 

eau fc Simard 1998 


Studies of active galactic nucleus (AGN) host galaxies give us the opportunity to study 

black holes ( 

Kormendv & Richstone 


Maeorrian et al. 


Gebhardt et al. 


McLure & Dunlop 


Marconi & Hunt 

20031 ). In cases where central black hole masses are independently determined, accurate 
decompositions of AGN host galaxies from their central point-source light contributions 
allow for the direct study of how galaxy light distribution relates to black hole properties. 
Alternatively, when time-intensive observations of black hole mass are not available, we can 
use established bulge-black hole relations to determine the black hole masses from bulge 

However, two-dimensional morphological fitting of AGN host galaxies is more compli- 
cated than that of "normal" (or "inactive" ) galaxies because of the presence of a central point 
source, which is often quite bright. In order to extract host galaxy properties from a source 
comprised of a host plus a central AGN, spatial resolution is critical. Thus the Advanced 
Camera for Surveys (ACS) on the Hubble S pace Telescope (HST) is the instrument of choice 

for many AGN h ost galaxy studies to date fISanchez et al.ll2004l ; lAlonso-Herrero et al.l 12008 
Ballo et allboOTh . 

Of the large multi-wave length surveys current ly available, the Great Observatories Ori- 
gins Deep Survey (GOODS; iGiavahsco et al.ll2004l ) provides some of the deepest multicolor 
ACS data. AGN identification is possible because of deep X-ray imaging with Chandra, 
as well as ground-based optical and infrared spectroscopic follow up. Spitzer data provide 
additional information on total light but insufficient spatial resolution to separate the galaxy 
from the active nucleus. We perfornied de tailed morphological analysis on the GOODS ACS 
data set using GALFIT (jPeng et al.ll2002l ). the results of which are presented in a forthcom- 
ing paper (B. D. Simmons et al. 2008, in preparation, hereafter S08). In order to understand 
the accuracy of those results - in particular, to probe the well-known effects of surface bright- 
ness dimming and dependence of physical resolution on redshift in the presence of a central 
point source whose size does not change with redshift - extensive simulations of host galaxy 
morphology are required. 


Previously, ISanchez et al.l (120041 ) simulated 1880 single-component host galaxies with 
point sources fainter than the host galaxy. Here we present a full treatment of over 50,000 
simulated AGN host galaxies in the redshift range 0.1 < z < 1.1, with both single-component 
and two-component bulge-plus-disk morphologies, and central point sources that are both 
brighter and fainter than the host galaxy. Results of these simulations are intended to inform 
data analysis of AGN host morphologies, to better infer intrinsic host galaxy shapes from 
fitted morphological parameters in the presence of a central point source. 

We discuss the data from which we draw our noise properties and simulated sample in 
Section 2. The detailed fit procedures and the simulation parameter space are presented in 
Section 3, and we assess the ability of our fitting procedure to recover accurate host galaxy 
and point-source parameters from GOODS-like images of AGN host galaxies in Section 4. 


2.1. HST ACS Data 

The GOODS fields each cover an area of approximately 10' x 16' with a total of 398 
HST orbits in both fields. The ACS has a resolution of 0".05 pixel"^, and observations were 
taken in the F435W, F606W, F775W, and F850LP ACS bands, hereafter referred to as 

V, /, and z \ respectively. D a ta acq uisition and reduction are detailed in iGiavalisco et al. 

( 120041 ) and iKoekemoer et al.l (120021 ). Each of the five epochs of data was processed via 
the basic ACS pipeline and a further processing task called multidrizzle, which finds an 
astrometric solution for all 5 epochs in order to correct for geometric dis tortion, and at the 
same time removes cosmic rays from the image (IKoekemoer et al.ll2002l ). The final images 
have resolution of 0".03 pixel"^, and the magnitude limits for extended sources are B < 28.4 
mag, V < 28.4 mag, / < 27.7 mag, and z < 27.3 mag. 

From the reduced data, we randomly selected 450 inactive galaxies with stellarity class 
less than 0.8 {i.e., objects that are not point sources) and magnitude z' < 24.0 mag for 
use in those of our simulations that include real data (§3.3). Since the majority of AGN 
host galaxies selected for morphological analysis in current surveys using HST have redshifts 
such that their J-band data lie in the rest-frame B band (0.575 < z < 0.9), we performed 
simulations using the /-band images and noise properties. 


2.2. Noise Properties 

For fully simulated galaxies (described in §3.4), noise appropriate to the GOODS- ACS 
fields was added. The task of simulating the noise in the GOODS-ACS fields is complicated 
by the dithering process, which correlates the noise among nearby pixels. We therefore use 
actual noise pixels (i.e., source-free pixels) from the final GOODS-N and GOODS-S images 
rather than statistical models. We sampled 150 sub-sections of the GOODS images that 
contained no sources, 75 in the North field and 75 in the South, for a total of over 2.5 million 
noise pixels. Each of these sub-sections was cut into tiles of size 50 x 50 pixels; this resulted 
in 1042 tiles. These were then used to create 500 noise images of the same size as our data 
images by arranging random mosaics of the 50 x 50 noise tiles. We have verified that these 
new noise mosaics have the same overall noise properties (distribution of pixel intensities) 
as the original noise-only subsections of the GOODS images (Figured]). 

We also tested noise mosaics created from 10x10 tiles of background noise but found that 
the distribution of pixel intensities was shifted compared to the correlated noise properties 
of the dithered ACS images. We thus used only the 50 x 50 pixel noise images to create our 
final noise mosaics. The noise value distributions of the 50 x 50 and 10 x 10 noise mosaics 
are compared to the actual GOODS noise distribution in Figure [1] 

3. Creation and Morphological Fitting of Simulated Samples 

We performed two distinct sets of simulations to test the accuracy of derived morpho- 
logical parameters. First, we used a sample of 450 real galaxies from the GOODS fields 
and added point sources to their centers. This allowed us to test the ability of GALFIT 
to extract point sources from the centers of real galaxies. It also informed our choice of 
parameters and galaxy types for the second, more extensive, set of simulations. In the latter 
set of simulations, we created a sample of 12,592 completely simulated GOODS galaxies, 
in order to additionally test the recovery of a wide range of host galaxy parameters. We 
fitted all of our simulated galaxies with our batch-fitting algorithm (described below). We 
also simulated the redshifting of these galaxies to quantify redshift-dependent effects, and 
we added noise taken directly from real GOODS data. 

3.1. Fit Procedure 

We performed morphological analysis in two dimensions usin g GALFIT, whic h can 
simultaneously fit an arbitrary number of components to an image fjPeng et al.ll2002l ). The 

- 5 - 

-0.02 0.02 

Pixel Value 

Fig. 1. — Histograms of pixel intensities for GOODS noise (solid), noise mosaics using a 
50 X 50-pixel sampling of the GOODS noise (dashed), and using a 10 x 10-pixel sampling 
of the GOODS noise (blue, dot-dashed). The histogram for the mosaic using 10 x 10-pixel 
tiles is significantly different from the intrinsic noise distribution. The distribution of pixel 
intensity of the 50 x 50 noise mosaic is nearly indistinguishable from that of the intrinsic 
GOODS noise distribution. 

- 6 - 

program uses a minimization method to determine the best-fit parameters. We used the 
Sersic profile, which models the light distribution of a galaxy as an exponential function 
with a varia ble half-light radius, r^, and an exponential parameter, n, called the Sersic index 

(ISersidll968l ). When n is fixed at a value of 1, the Sersic profile is equivalent to an exponential 
disk; when n = 4, the Sersic profile is equivalent to a de Vaucouleurs bulge. Each of our 
simulated galaxies was fitted twice: once using a single Sersic function with a variable index, 
and once with two Sersic functions with fixed n = 4 and n = 1 (bulge -|- disk). We also 
simultaneously fitted point-source (PS) components using the point-spread function (PSF) 
of the data, determination of which is based on GOODS field stars and is described in SOS. 

We developed a batch-fitting algorithm to fit AGN host galaxies in GALFIT (described 
in detail in SOS), which we use here on our simulated galaxies. The batch-fitting algorithm 
uses a series of initial guesses to execute a first-pass fit of the central region of each AGN host 
galaxy, fitting the individual centroid positions of each component (Sersic plus point source, 
or de Vaucouleurs bulge plus exponential disk plus point source). The second fit iteration 
zooms out to include the full extended galaxy and fits the central point-source magnitude and 
AGN host galaxy parameters (Sersic index or bulge/disk decomposition, half-light radius, 
axial ratio, etc.). Finally, a third iteration is performed where all parameters are allowed to 
vary (except n in cases of a bulge -|- disk fit). 

It is important to note that, while these simulations include thousands of AGN host 
galaxies, even the largest (current) galaxy surveys using ACS data provide no more than 
hundreds of AGN host galaxies that can be reasonably fitted with GALFIT. We can therefore 
(as in SOS) follow up the batch fits by hand, using the results of the batch-mode fitting as 
initial guesses to further constrain the results for each AGN host galaxy. Due to the size of 
the simulated galaxy sample in this paper, we did not do this final hand-fitting step. Thus 
our conclusions about the fraction of cases for which accurate morphological parameters are 
recovered should be considered conservative when compared to a sample of galaxies whose 
fits are adjusted and verified individually. 

3.2. Determination of Initial Parameter Guesses 

For our real host galaxy sample, we use magnitude, flux radius, and position angl e from 
the GOODS catalogs, which were created using SExtractor (IBertin &: Arnoutslll996l ). Our 
simulated galaxies do not have a SExtractor catalog, but we simulated initial parameter 
guesses by introducing random errors into the true values. 

Specifically, when determining initial guesses for parameters that can be measured from 

- 7- 

a data image, we assume that the guess value is accurate to within the following: ±(2.0, 2.0) 
pixels in position, ±0.5 in total magnitude, 10% in and b/a, and 10° in position angle. 
We assumed no a priori knowledge of the Sersic index n of each simulated galaxy, nor of 
whether each simulated galaxy was a two-component bulge plus disk system, or a system 
with either a pure bulge or disk. Our initial guess for the Sersic index (or indices) of each 
fit is n = 2.5. 

In the case of a multi-component fit, we assume that the total magnitude is split evenly 
among the components (i.e., that we have no prior knowledge of host:PS contrast ratio or 
bulge-to-total ratio), and we additionally assume that the measured position angle and axial 
ratio are the average of the actual values for each separate component after adding the 
random fluctuation to the sum. 

3.3. Real Galaxies with Added Point Sources 

In order to test whether we can recover central point sources from galaxies, we selected 
450 J-band images of normal galaxies (i.e., not X-ray detected) from the GOODS-North and 
GOODS-South fields. The selection was performed so the sample has the same magnitude 
distribution as the full sample of galaxies in each of the GOODS fields, but was otherwise 
random. We assumed no knowledge of redshifts within the sample. 

We fitted each of these galaxies with a single Sersic profile in order to determine the 
baseline set of morphological parameters for each galaxy. These fit parameters (Sersic n, 
magnitude, half-light radius re, etc.) were taken to be the "actual" galaxy parameters. 

We then made nine copies of each galaxy, adding one point source to each, centered on 
the central pixel of the galaxy. The added point sources ranged in /-band magnitude from 
21 to 29, in increments of 1 mag. This gave us a total of 4050 simulated AGN from the 
original 450 galaxies, with a large range of host galaxy-to-point-source contrast ratios, from 
~1000:1 to 1:100. These provide a straightforward means of assessing the effect of a central 
point source on the recovered parameters of a real galaxy. 

Our convergence rate for the fitting routine depends on the magnitude of the added 
central point source. Of the models with Ips = 21, 88% converge, whereas 64% of models 
with Ips = 29 (significantly below the flux limit of our J-band sample) converged. This is 
due to the fitting program being unable to converge to a value for the central point-source 
magnitude. Examples of fit results for typical galaxies in our sample are shown in Figure [2j 

- 8 - 

Fig. 2. — Example fit results for two AGN hosts simulated from real galaxies. For each galaxy, 
the top row shows the original GOODS galaxy (from left to right: galaxy, fit, residual). 
Each row below shows the host+fit+residual for added central point sources of magnitudes 
Iab = 22, 24, 26, and 28, with the faintest point source in the bottom row. Note: These 
cutouts are zoomed-in to show the galaxies; the actual fitting regions are significantly larger. 

- 9 - 

3.4. Simulated Galaxies 

The simulated hosts described in §3.3 provide information on how the presence of a 
central point source changes the fitted host galaxy parameters, but they do not allow us to 
explore redshift effects or quantify the absolute accuracy of the fit parameters. This can 
only be accomplished with a set of simulations for which we know the intrinsic properties of 
each simulated galaxy. 

To that end, we simulated three kinds of galaxies with a range of parameters: pure de 
Vaucouleurs bulge (B) galaxies with fixed Sersic n = 4, pure exponential disk (D) galaxies 
with n = 1, and composite galaxies with both bulge and disk (B+D) components in varying 
proportions. The galaxy parameters were chosen to be typical of bright, local galaxies 

( iBinney fc Merrifieldl Il998l ) and placed a.t z = 0.125. Redshifted samples were developed 

from these, as described below. 

For the B and D type galaxies, the component parameters occupy a grid of four values 
for each parameter across the ranges 16.9 < Iab < 20.4 and 1.5 < < 6.0 kpc (for bulges); 
4.0 < Te < 10.0 kpc (for disks), with axis ratios 0.25 < b/a < 1.0. For the single-component 
fits, the position angle was fixed at 45.0°. This results in 64 B and 64 D galaxies, each 
of which is "inactive", i.e., without a central point source. We create simulated AGN host 
galaxies by adding point sources to these. We used five values between 16.5 < Iab < 24 
for the central point source, resulting in 320 B-type and 320 D-type AGN host galaxies, 
with contrast ratios ranging from —1.3 < log (-^hostZ-^PS) — Lastly, in order to ensure 
enough galaxies for statistical analysis after binning the fitted single-component sample, we 
made four copies of each B and D inactive galaxy and two copies of each B and D AGN host 
galaxy. This resulted in a total of 256 B and 256 D inactive single-component galaxies, and 
640 B and 640 D AGN host galaxies. 

For the B+D type galaxies, the parameter space includes more magnitudes so as to pro- 
vide 20 input bulge-to-total ratios with 0.028 < (B/Tot) < 0.97 [plus the single-component 
B and D fits, which have (B/Tot) = 1 and 0, respectively]. In addition, the position angles 
of the individual components were allowed to vary such that some of the simulated B+D 
galaxies have components that are slightly (< 15°) off-axis with respect to each other. The 
axis ratios between bulges and disks were also allowed to vary with respect to each other. 
These changes significantly increase the total number of double-component galaxies created; 
thus the number of radius, b/a, and point-source magnitude parameters used was decreased 
in order to keep the total number of galaxies from being prohibitively large from a computa- 
tional perspective. This resulted in the creation of 2700 inactive double-component (B+D) 
galaxies, and 8100 B+D AGN host galaxies. Table [T] gives the parameter values used to 
create the entire local suite of 12,592 simulated single- and double-component galaxies. 

- 10 - 

Since we are also interested in distinguishing the effects of redshifting galaxies from the 
effects of evolving galaxies, the initial sample was defined to lie at z = 0.125, and three 
additional samples were placed at z = 0.413, 0.738, and 1.075, corresponding to the redshifts 
at which the centers of each of our GOODS filters (F435W, F606W, F775W, and F850LP) 
are in the rest-frame B (F435W) band. 

To redshift each galaxy, we used a concord ance cosmology with fltot = 1, = 0.73, 

and Hq = 71 km s ^ Mpc ^ (jSpergel et al.ll2003l ) to calculate the cosmological dimming and 

loss of resolution corresponding to each redshift. Because the size of the ACS PSF does not 
change with redshift, we assume that the size of AGN central point sources will also stay 
fixed with redshift, unlike the host galaxy. We therefore created the redshifted AGN by 
redshifting the model host galaxy (in flux and size) separately from the central point source 
(in flux only) and adding them together before convolving with the ACS PSF and adding 
noise to the image. The redshifting of each of the 12,592 galaxies produced another 12,592 
galaxies at each final redshift, for a total of 50,368 fully simulated galaxies located at four 
different redshifts from 0.125 < z < 1.075. 

We fit each using our batch-fitting algorithm. Each galaxy was fit twice: once with a 
generalized Sersic profile plus a point-source component, and once with a combination of 
de Vaucouleurs bulge, exponential disk, and point source. Inactive galaxies were also fit 
without point-source components. Examples of fit results for typical galaxies in our sample 
are shown in Figure [31 

4. Results And Discussion 

4.1. Existing Galaxies with Added Point Sources 

We are interested mainly in three properties of these models: how well they recover the 
input point-source magnitude, how well they recover the baseline host galaxy parameters, 
and how the accuracy of the host galaxy parameters relates to the point-source magnitude 
and/or the contrast ratio between central point source and host galaxy. 

4.I.I. AGN (Point Source) Recovery 

Figure H] shows the degree of recovery of the central point-source magnitude for input 
point sources. As expected, brighter point sources are more accurately recovered. At all 

- 11 - 

i = 0.12S 0J12S 0.737S 1JI7S 

HostPS 1 






♦ : 








Fig. 3. — Example fit results for four fully simulated galaxies at four different redshifts. Each 
row shows one galaxy. The first two columns show the image of the galaxy aX z = 0.125 and 
residual. Each successive pair of columns from left to right shows galaxy and residual for 
z = 0.4125,0.7375, and 1.075, respectively. Note: These cutouts are zoomed-in to show the 
galaxies; the actual fitting regions are significantly larger. 

- 12 - 

values of input point-source magnitude, the fitted value of the point-source magnitude tends 
to either converge to the true input magnitude or remain at the input guess magnitude. 

In all cases, we can easily separate these two groups with a simple reduced cut 
(x^ < 2.0). After the cut, the number of remaining poor fits is very low, as is the number 
of good fits removed by the cut: both are less than 5%, even in the limiting cases of very 
high and very low input point-source magnitude. This level of contamination/excess removal 
of good fits is not strongly dependent on the value at which we choose to cut nor is it 
strongly dependent on the input point-source magnitude. 

The rms uncertainties of the recovered point-source magnitudes for each input point- 
source value are shown in Table O These values reflect the systematic uncertainties in 
the fitted parameters, and should be added to the statistical error in fitted point-source 
magnitude returned by GALFIT. 

4.. 1.2. Host Galaxy Parameter Recovery 

Figure [5] shows the correlation between the fitted model's host galaxy magnitude and 
the actual host magnitude (as determined before the addition of the central point source). 
The percent completeness of the low-error sample compared to the full sample varies greatly 
across a range of contrast ratios between host galaxy and central point source. For very 
faint central point sources, the host galaxy magnitude is very well recovered, whereas for 
very bright point sources the fraction of host galaxies with well-recovered parameters is very 
low (~ 15%). 

Our simulations show that the reduced parameter is generally not sufficient to dis- 
tinguish good from poor host galaxy fits. If we also limit the relative error of the effective 
radius parameter, however, the fraction of selected galaxies that recover the true magnitude 
of the host galaxy improves significantly. With the combination of x^ and Ur^/'^e parameters, 
we remove more than 85% of the fits with magnitude differences of 0.5 dex or more from the 
actual galaxy magnitudes. The number of well-fitted galaxies that are also removed from 
the sample ranges from 5% to 20%, depending primarily on how conservatively we choose 
our relative 0"^^ cut. Figure [5] reflects a cutoff value of (cXr^/re) = 0.8, which removes 88% of 
poor fits and 12% of the good fits from the /pg = 24 sample. 

Figure [6] shows how well we recover the Sersic index, n. The same error cuts (x^ and 
^Tel'^e) efficiently reject the inaccurate fits. The fraction of models whose fitted n values 
deviate by more than An = 2 compared to the n values of the unaltered galaxies increases 
significantly with increasing point-source fiux. That is, a very bright central point source 

- 13 - 








1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 

Input nipg = 22 " 
N„„, = 389 



o o 


1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 

Input nipg = 24 " 
N,„„„ = 394 


O o 

o o 

- o ° o ^ _ 

- X — o— ^ x-i^.^aC*:!«t^»i|&^Hl^^^ 

' 1 1 1 1 1 1 1 1 1 1 1 1 1 

' 1 1 1 1 1 1 1 1 1 1 1 1 1 1 

Input nipg = 26 " 
N,„„, = 372 



X - 

X ^ 

— o ^ o ^ _ 

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 

X Input mpg = 28 " 

: = 315 : 


X X 
X X J! 

X X >^ >t''>V<x 

O O O X *x 

X ^ O '^rtS^--'' 


1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 

20 22 24 26 20 22 24 26 

PS initial guess magnitude 

Fig. 4. — Recovery of magnitude of a point source added to real galaxy images, for: input 
point-source magnitudes of 22, 24, 26, and 28 mag. The fitting recovers the point-source 
magnitude nicely after making a simple xt < 2.0 cut (black crosses), with a larger spread for 
fainter values of the input point source, as expected. Fits with a large value of xl are shown 
as green open circles; Nconv indicates the number of fits (out of 450) that converged automat- 
ically for this input point-source magnitude. The dashed horizontal line shows m^^^ = m^^, so 
fits falling on this line perfectly recover the actual point-source magnitude. The dotted line 
shows m^^t = mguess: fits on this line converged to their initial guess point-source magnitude, 
and the final fitted magnitude is wrong; nearly all of these fits are removed by the xi cut. 








1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 

Input m„„ = 22 

_ % PS 

o N,_^= 389 

n X' ~ 

A ^^l 
'^A O >< 

,•' A 

^^^^^ ' 

^,*A ^ o ~oo 

X .•• A o 
O A 

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 

Input mpg = 24 
N = 394 



x^ ^f^^^^^ 

- X o ^r?4^^i 

O- X ^x^ *.A A ^ 

X X ^ ^ -A^° A 

X. A A^ 
A A ^ 

- A^ 

1 Aa\'~I\^! I 1 

•"l 1 1 1 1 1 1 1 1 1 1 1 1 1 ,. 

Input mpg = 26 
~ ' = 372 ^ ^x-'' : 

- xJ^* 

-_ - 

Xo ''^X^^'^.XA* ^ 

x°' x^k' 

X X A O ^ A 

- - n " 

O,-- A AAA 

- .' A — 
O.-' A 

^ A 

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 

• 1 1 1 1 1 1 1 1 1 1 1 1 1 .. 

Input mpg = 28 

N =315 A-'' 

„ conv !|( 
J^aX^ a 

X ^ ^X ^ A 

° "i:-^^ x" A o 

^ X AAA 

X • 


1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 

20 22 24 26 20 22 24 

Galaxy Magnitude 

Fig. 5. — Fitted host galaxy magnitude versus actual galaxy magnitude for fits to actual 
galaxies with added point sources. Excluding high xl values and effective radii with large 
fractional errors ensures a high fraction of good fits, with moderate uncertainties. Green 
open circles indicate fits with high xt values. Cyan open squares indicate fits with excessive 
errors on the effective radius fit parameter (ar^ > 0.8 x rg). Purple open triangles represent 
fits with both high xl and high a^.^. The dotted line represents equal input and recovered 
magnitudes (i.e., perfect recovery). The error cut removes some of the fits that do lie on the 
1:1 line, regardless of the precise value for the cut. 

- 15 - 

can easily be confused with a concentrated (high-n) Sersic value. For all values of the 
added point-source magnitude, however, the same combination of xl and relative r^, error is 
sufficient to remove 88% of the inaccurate fits from the data, while removing less than 15% 
of the well-fitted models. 

4.1.3. Dependence of Parameters on HostiAGN Contrast Ratio 

Figure [7] shows the relation of fitted host galaxy parameters to the contrast ratio between 
the host galaxy and the added central point source, for the 1491 (of 4050) sources that are 
not eliminated by the combination of and (Tr^/rg cuts. These include galaxies with all 
possible values of input point-source magnitude. 

The left panel of Figure [7] shows the difference between the fitted host galaxy magnitude 
and original galaxy magnitude for the PS-added models. The recovered magnitudes are no 
different on average than the correct parameters. For galaxies with contrast ratios of at 
least log (Lj^Qg|./Lpg) < —0.6, the host galaxy magnitude is well recovered, such that a 
weighted least-squares fit to those data with contrast ratios (Lj^Qg|-/Lpg) greater than 1:4 
has a slope within 1% of zero, indicating no correlation between contrast ratio and recovered 
host galaxy magnitude. The scatter in the plot indicates that, while the sample as a whole 
is well-recovered, individual galaxies may have large differences in recovered parameters due 
to the presence of a point source. We see this result again in the right panel, which shows 
the difference between fitted and original Sersic index as a function of host-to-PS contrast 
ratio. The inaccuracy in Sersic index is larger when the added central point source is more 
than 4 times as bright as the host galaxy: the slope of the weighted least-squares fit is 6% 
for ^Iab and 11% for An. With proper error cuts, and for galaxies brighter than 1/4 of 
the point-source magnitude, our host galaxy fits are not significantly contaminated by the 
central point source. However, the presence of a point source increases the uncertainty in fit 

Figure [S] shows the median change in fitted parameters with input point-source magni- 
tude. For very bright added point sources {Iab = 21, (Host:PS) ~ 1 : 11), all fit parameters 
deviate from their initially-determined values. The median Sersic index is low by approx- 
imately An = 1.75, indicating that on average, galaxies with such bright point sources 
could be classified as disk-dominated even if they have more bulge-dominated intrinsic mor- 
phologies. The distribution is also very wide in the brightest-point-source bin: for such low 
host-to-PS contrast ratios, the Sersic index is essentially completely uncertain. For fainter 
point-source magnitudes, recovery of the Sersic index is far more reliable, although the un- 
certainty (indicated by the width of the distribution in each bin) is considerably larger than 

- 16 - 




n — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — r 

Input rripg = 22 
N.„„„ = 389 







10 - 


Input m 



^ 20 h 



N.onv = 372 



O §L ^ ^ 


n — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — r 

Input nipg = 24 
N^„„„ = 394 


Input m 



= 28 


OOo o 

o °@ 

O OA^ 

O '~'a a a X 
O OoO^^^>^ ^ 

<^ O - x^ >«< .: 


J I I I I I I e_j I I I I I L_ 

20 22 24 26 28 20 22 

Host Magnitude 




Fig. 6. — Change in Sersic indices of host galaxies after adding point sources with original 
host magnitude. Excluding high xl and (Tr^/re yields the correct Sersic index (a proxy for 
morphological type). The color-coding of the points is the same as in Figure (5) green open 
circles for high x^, cyan open squares for high relative ar^, and purple open triangles for 
both high xl and relative cr^^ . The dotted lines at An = indicate perfect recovery of the 
Sersic index. As expected, the bright-point-source fits return mostly unreliable host galaxy 
parameters, in contrast to the faint-point-source fits, which return mostly reliable host galaxy 






Fig. 7. — Accuracy of the host galaxy magnitude (left) and Sersic index (right) as a func- 
tion of the ratio of intrinsic galaxy luminosity to the added point-source luminosity. In 
both figures, contour plots are overlaid on the data points. Weighted least-squares fits to 
all data points (dashed lines) indicates that the recovery of galaxy magnitude and Sersic 
index is very slightly correlated with host-to-PS contrast ratio. Excluding points with 
log (Lj^Qg|./Lpg) < —0.6, the weighted least-squares fit (solid lines) is consistent with zero 
slope (i.e., no correlation between Host-to-PS contrast ratio) for both plots. 

- 18 - 

that reported by the fitting routine. 

The ratio of host and point-source luminosities is also incorrect at the brightest point- 
source magnitudes: in these situations the typical fitted point-source magnitude is very close 
to the input value, and the fitted host magnitude is too low (bright) by approximately 1.45 
dex. Interestingly, the deviations in magnitude do not add to zero: because bright point- 
sources are also slightly too bright, the total magnitude is also underestimated (too bright) 
in the presence of a very bright central point source. 

For point sources brighter than Iab = 24, the distribution of recovered host galaxy 
magnitude is wide, such that the 68% confidence intervals span more than 1 dex in all bins. 
For those point sources with magnitudes fainter than 24 (which is approximately equal to 
the average magnitude of our initial galaxy sample), on the other hand, the model's host 
galaxy magnitude is well recovered. 

Since the faintest point-source magnitudes {Iab = 28, 29) are below the 10a point- 
source detection limit of our band, the fitted point-source magnitude for these cases is often 
at least 1 mag brighter than the input point-source magnitude, with 68% confidence intervals 
larger than 1 mag. 

The Sersic indices for the host galaxies are within An = 0.25 on average for most values 
of the added point-source magnitude. Since disks are generally more extended than bulges, 
the Sersic parameter is lower (on average) when the central point source is so bright that the 
central portion of the host galaxy is dwarfed by the light in the wings of the central point 
source. In all but the brightest cases, however, the median difference between the model 
Sersic index and the no-point-source inde x is smaller than the si ze of the morpholo gical bins 

discussed for a typical AGN sample (e.g.. ISanchez et al.l (120041 ): iBallo et al.l (l2007l ). S08), so 

it is unlikely that a significant fraction has been mis-classified as disk-like or bulge-like due 
to the presence of a central point source. 

The rms values of the recovered Sersic indices and recovered host magnitudes for each 
input point-source magnitude are given in Table [21 

4.2. Fully Simulated Galaxies 

4-2.1. Input Parameter Recovery 

For our 12,592 fully simulated GOODS galaxies, we apply the error cut determined 
above, using a combination of xl (^re/^e to distinguish good fits from poor fits. Within 
the single component galaxies, this error cut results in 706 of 896 B type galaxies with "good" 

- 19 - 

t ^ 

1 I 

'\ I 


24 26 
input PS magnitude 


Fig. 8. — Median change in Sersic index n (upper), host magnitude (middle), and point- 
source magnitude (lower) versus input point-source magnitude. Each point represents the 
median value of all the galaxies with the given input point-source magnitude whose fits pass 
the Xu (Tr^/re cut described in the text. The host Sersic indices (top panel) are generally 
well-recovered for all but the brightest input point sources. The host magnitudes (middle 
panel) are also well-recovered in all but the brightest input point-source bin, with the error 
bars (representing the width of the distribution encompassing 68% of sources) increasing with 
increasing point-source strength. Point-source magnitudes (bottom panel) are recovered well 
until the input magnitude is fainter than the published lOa point-source detection limit of 
the data. 

- 20 - 

fits and 708 of 896 D type galaxies witli "good" fits. 

The recovery of tlie Sersic index for both of these types of galaxy models is shown for 
the z = 0.125 sample in Figure O If we classify galaxies as bulge-dominated when their 
fitted n > 2.0 and disk-dominated when their fitted n < 2.0, the level of misclassification 
based on these simulations is less than 15% for AGN host disk galaxies and less than 10% 
for AGN host bulge galaxies. Fewer than 10% of inactive disks are misclassified as bulges 
using the above cut; no inactive bulges are misclassified as disks. For any local sample of 
single-component galaxies (the simplest possible case), then, the Sersic parameter is a very 
reliable indicator of morphological type. 

The recovery of the z = 0.125 sample of two-component (B+D) galaxies is shown 
in Figure [TUl The fitted bulge-to-total light ratio, B/Tot = Lb/{Lb + L£)), is within 
I A (B/Tot)| < 0.15 for 50% of our AGN host sample and 66% of our inactive sample. A 
Gaussian fit to the data in Figure [10] gives an estimate of uncertainty in the bulge-to-total 
parameter of ae/Tot = 0.04. However, the contamination from very deviant fits (i.e., the 
wings of the histogram in Figure [TUi) is large. Approximately 14% and 13% of host and 
inactive galaxies remaining in the sample after our error cut have A (B/Tot) < —0.15, and 
36% of host galaxies and 21% of inactive galaxies have A (B/Tot) > 0.15. 

Figure [11] shows the ratio of fitted-to-input bulge-to-total flux ratio as a function of the 
intrinsic contrast ratio between the host galaxy and the central point source. The median and 
la error bars (defined here as the values encompassing 68% of the points in each bin) for the 
distribution of simulated galaxies without central point sources are plotted for comparison 
at a value of Lhost/Lps = 3. In general, the presence of a central point source skews the 
fits to slightly high values of the bulge-to-total ratio. For galaxies where the central point 
source is very bright, the spread is higher, such that a galaxy may fit to a bulge-to-total ratio 
twice the intrinsic value and still be within la of the median value. Based on this, one could 
attempt to correct the bulge-to-total ratio for the presence of a central point source, but a 
bulge-to-total ratio fitted to a host galaxy could be overestimated by as much as 100%. 

This large tail of high fitted bulge-to-total values has several causes. The radius of the 
bulge may converge to an unphysically high value; this causes the luminosity bulge-to-total 
ratio to skew high. This appears to be slightly more common when the intrinsic luminosity 
of the galaxy is disk-dominated, i.e., with (B/Tot) < 0.3. In addition, if the total source 
magnitude is not well-conserved during the fit, the bulge-to-total ratio typically deviates 
from the input value by at least 10% (high or low). This is more likely to occur when the 
bulge and disk components have approximately equal luminosity, or the bulge is slightly 
brighter [0.45 < (B/Tot) < 0.7]. 

- 21 - 


150 - 

"1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1- 

451 Bulge AGN 

453 Disk AGN 








r I 
i "i 

_l I I I I I L_ 

_l I L_ 

255 Bulges (no AGN) 
255 Disks (no AGN) 

I "I 


_l I I I I I I I I I I I I I L 

_l I L_ 

4 6 
Sersic n 


Fig. 9. — Recovery of Sersic index n for the single-Sersic fits to simulated single- component B 
and D AGN host galaxies (blue, top) and inactive galaxies (black, bottom). For the bulge- 
dominated B galaxies (solid histograms), the average fitted Sersic index is slightly higher 
than the input n = 4 in both AGN and inactive galaxies; however, these are unambiguously 
classified as bulge-dominated galaxies, with fewer than 10% of bulge host galaxies mis- 
classified as disk-dominated galaxies and no bulge inactive galaxies mis-classified as disks. 
For the disk-dominated D galaxies (dot-dash histogram), the recovery of the input n = 1 is 
also excellent, with 13% of D host galaxies and 9% of D inactive galaxies having n > 2.0. 
The separation between the typical fitted n for B and D galaxies is robust and indicates that 
we can use the Sersic index as a strong indicator of galaxy morphology for single-component 
local galaxies with and without central point sources. 

- 22 - 

400 - 




Fig. 10. — Recovery of input bulge-to-total ratio for the fully-simulated two-component 
galaxies. We define A(B/Tot) > to mean the fit was more bulge-like than the input 
galaxy. A Gaussian fit to the Host galaxy distribution gives asT = 0.04, but there is a 
large asymmetric tail on each distribution indicating that the samples are skewed toward 
more bulge-like fits than the intrinsic distribution. 36% of host galaxies and 21% of inactive 
galaxies have A (B/Tot) > 0.15. 

- 23 - 

3 - 

2 - 



1 - 

o o 



_J I I I I \ I I 


Input log(Lj^^^yL 


Fig. 11. — Ratio of fitted-to-input bulge-to-total vs. input host-to-PS contrast ratio. Points 
plotted are the median values in each bin, with error bars marking the range of values 
containing the central 68% of the points in each bin. The open circle shows simulated 
galaxies with no added point source; this bin recovers input bulge-to-total ratio closely, with 
a tail of high bulge-to-total values. For simulated host galaxies, the recovered bulge-to-total 
ratio is generally high by 10-20%, and the distribution of points in each bin is wider for 
galaxies where the central point source is at least as bright as the host galaxy. 


About 18% of the two-component AGN host fits result in either a bulge or disk half- 
light radius that is unphysically low given the input sizes of both in our sample. These fits 
span the entire range of A (B/Tot) values, and are far more common when the central point 
source is intrinsically bright. For the locally-defined sample, 79% of B+D bulge components 
and 63% of B-l-D disk components with fitted re < 5.0 pixels are those with the brightest 
central point source; this comprises 26% of all two-component sources with the brightest 
central point source. It is worth noting that these fits typically show residuals that clearly 
indicate a poor fit, and they can therefore be removed from the well- fitted sample on visual 
inspection, or followed up with individual fitting to improve the fit. 

The recovery of bulge and disk sizes in our sample is shown in Figure[T2l The distribution 
of Are for single-component galaxies (pure bulges and disks) is sharply peaked at a value of 
Ar ^ (in pixels). A total of 66% of pure bulges and 65% of pure disks recover their input 
radii within — 1 < Are < 12 pixels. The discrete peaks at negative Are represent those fits 
that converged to unphysically low values (re < 5.0 pixels for the local sample). Less than 
5% of single-component fits converged to extremely large radii (Are > 150 pixels). 

The radius parameter converges to extremely high values far more frequently for the 
two-component fits, and with higher frequency for the bulge components of two-component 
galaxies (20%, B+D bulges) than for the disk components of two-component galaxies (13%, 
B+D disks). The high- Are tail for B+D bulges is a result of confusion with the central point 
source of the simulated AGN host. Not only are bulges more compact than disks within 
our B+D galaxies (and typically within real galaxies as well), but the slope of their light 
profile more closely resembles the light profile of the ACS PSF than does an exponential disk 
profile. This can lead to uncertain bulge radii and magnitudes, especially when the intrinsic 
luminosity of the bulge is similar to that of the central point source. Disk components of 
B+D galaxies are not as readily confused with the central point source: their fit typically 
fails when the surface brightness of the disk component is too low compared to the bulge 
and point source components. 

Figure [T3l demonstrates a possible method of separating good from poor two-component 
fits. After removing those fits where either the bulge or disk converged to < 5.0 pixels, 
additionally requiring that the total source magnitude be conserved to within |Amto(| < 0.5 
(determined by comparing the fitted total source magnitude with a measured, aperture- 
corrected magnitude), and/or that the flux- weighted effective radius of the fit be no more 
than 90 pixels larger than the measured half-light source radiu^ removes 81% of sources 
with I A (B/Tot)| > 0.15 and retains 83% of sources within 0.15 of the input bulge-to-total 

This value was chosen because it is twice the observed radius of the largest source in our sample. 

- 25 - 


"T — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — r 

- 451 Pure Bulges 





I I I I I I I I I I I I I I I I I I I I I I I I I I 1 . 

6818 B+D Bulges 

0.3 - 



1 — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — I — r 

453 Pure Disks 

I I I I I I I 

I I I I I I I I I I I I I I I I I l _ 
6818 B + D Disks 

100 100 200 300 -100 100 200 300 

Ar [Output — Input] 

Fig. 12. — Histograms of the change in half-hght radius for pure bulge (top left), pure disk 
(top right), B+D bulge (bottom left) and B+D disk (bottom right) fits to simulated AGN 
hosts. The majority of fits converge to a value that is equal to or greater than the input 
value of the half-light radius for that source or source component. The small discrete peaks 
at negative Ar^ values are comprised mostly of fits for which re converged to an unphysical 
value for our sample, < 5 pixels. Single-Sersic fits typically recover the sizes of pure 
bulges and pure disks with about the same level of success. However, the local sample of 
two-component galaxies indicates that bulge component radii are more likely to converge 
to a value that is at least 50 pixels too high, or at least twice the intrinsic radius. Local 
disk components are more likely to be underestimated, with a lower a for the high-Arg 
distribution than for B+D bulges. 

- 26 - 

— I 1 1 1 1 1 1 1 1 — 

X Ar^ ^ 90 pixels, |Am^„J ^ 0.5 

c Ar^ g 90 pixels, |Am^„J > 0.5 

□ Ar^ > 90 pixels, |Am^^J g 0.5 

Ar^ > 90 pixels, lAm^^^J > 0.5 
* Tg g 5.0, g 5.0 pixels 

I I I I I I I I I I I I I I I I I I I I L 

-1 -0.5 0.5 1 

A(B/Tot) [Output - Input] 

Fig. 13. — Deviation from input host-to-PS contrast ratio versus deviation in bulge-to- 
total ratio. Fits located at (0, 0) perfectly recovered the input bulge-to-total ratio and the 
Galaxy- AGN luminosity ratio (Lj^Qg|./Lpg). Requiring that the difference between the half- 
light radii of the host be no more than 90 pixels larger than that observed for the entire 
source (Arg < 90), that the total source magnitude luminosity be conserved to better than 
|Amfoi| < 0.5 mag, and that the radii of neither the bulge nor the disk component be less 
than 5 pixels (typically indicative of a poor fit) cuts 81% of sources with |A (B/Tot)| > 0.15, 
and retains 83% of sources with |A (B/Tot)| < 0.15. 


ratio. This also constrains the fitted host-to-PS contrast ratio to within 8% of the input 
value for 95% of the remaining "good" sample. However, this selection biases the remaining 
sample toward sources with faint central nuclei and preferentially retains bulge-dominated 
sources with intrinsic (B/Tot) > 0.7. In general, the result of two-component fits should be 
used with caution, even for a local, well-understood sample. 

4-2.2. Detection of Central Point Sources 

We established in §4.1.1 that central point sources can be recovered on average for a 
sample of simulated host galaxies created from the addition of a central point source to a 
real GOODS inactive galaxy, although the fraction of recovered point sources varies depend- 
ing on the luminosity of the added point source. Using the xl> cut previously determined 
from this sample, we now examine the recovery of point sources with redshift and intrinsic 
morphological type. Figure [H] shows the fraction of fits in which the central point source 
was detected for our entire sample of fully simulated galaxies. In this case, our criteria for 
a "detected" point source is simple: the point source is considered undetected only if the 
magnitude parameter converged to a value consi derably fainter than the detection limit of 

our simulated data [rnps = 27.1 for our sample; iGiavalisco et al.l fl2004l )). Thus Figured 
shows the fraction of fits for which any point-source magnitude was recovered, rather than 
those fits for which the magnitude is close to the intrinsic magnitude. 

Figure [H] indicates that the point source is detected within the fit at least 90% of the 
time for all values of input log (Lj^Qg|-/Lpg). The fraction is lowest when the point source 
and host galaxy have about equal luminosity. The detection fraction increases to more than 
95% at the highest values of input log (Lj^Qg|./Lpg), even for cases in which the point source 
is 100 times fainter than its host galaxy. This is due to the fact that the convergence fraction 
is lower for these bins; fits where the central point source may not have been detected are 
less likely to converge, driving the detection fraction higher for those fits that did converge. 

While we are primarily interested in the recovery of host galaxy parameters and optical 
detection of central AGNs, we also want to characterize the fraction of spurious optical 
point-source identifications. Thus, Figure [T5l shows the results of a Sersic + PS fit to inactive 
galaxies with no intrinsic central point source. We find a significant difference in the number 
of false point-source detections between pure disk and pure bulge galaxies. Only ~ 1% of 
pure disk inactive galaxies detect^a central point source when none are present. However, we 

^We define "detected" as having been fit with a magnitude brighter than the pubhshed lOcr point-source 
detection limit of the GOODS survey (from which we take our noise properties). 

- 28 - 

Fig. 14. — Fraction of point sources recovered for single-component fits to all simulated 
AGN host galaxies. The fraction of recovered point sources is slightly lower when the with 
input log (Lj^Qg|./Lpg) is near zero {Lps ~ Lhost or slightly fainter). However, even in these 
cases, the fits detect at least 90% of point sources. 

- 29 - 

find a point source nearly one-quarter of tlie time wlien tlie simulated galaxy is a pure bulge. 
Because our total sample is composed of equal numbers of pure bulge and pure disk galaxies, 
this gives an overall false detection of a central point source in 12% of inactive galaxies. 
However, this effect is strongly dependent on morphology: fits to elliptical galaxies are far 
more likely to detect a faint point source when none are present than fits to disk galaxies. 
Host galaxy studies estimating the uncertainty in the fraction of hosts with detected optical 
point sources should therefore consider the morphological composition of their data sample: 
if the sample contains mostly elliptical galaxies, the number of false detections could be as 
high as 25%. 

4.2.3. Redshift Effects on AGN Host Galaxies 

At higher redshifts, surface-brightness dimming causes loss of extended components, and 
an unresolved central AGN corresponds to a larger fraction of a galaxy's central region at high 
redshift fo r fixed spatial resolut i on. Because our GOODS data sample (SOS), as well as those 

of ot hers fiSanchez et al.l 120041 : iGrogin et al.l 120051 : iBallo et al.l 120071 : lAlonso-Herrero et al. 
20081 ). extends to 2; ~ 1, we consider the quantitative effects of redshift on morphological 

After applying data cuts requiring Arg < 90 pixels and |Amtot| < 0.5 to our results at 
all redshifts, we assess the reliability of using the Sersic index to classify single-component 
galaxies at different redshifts in Figure [T6l We have "classified" a galaxy as bulge-dominated 
if it has n > 2.0 and disk- dominated if it has n < 2.0. Figure [16] then shows the fraction of 
correctly classified B- and D-type galaxies at each redshift. In general, the classification of 
pure bulges and pure disks is highly reliable for AGN host galaxies and point sources. At all 
redshifts, disks and bulges are correctly classified at least 90% of the time using these data 
cuts. Without these cuts, the misidentification rate increases to as much as 15%. 

The average recovery of input half-light radius for all our simulated galaxies (pure bulges 
and disks as well as B-l-D bulges and disks) is shown in Figure [T71 Each point represents the 
peak of the distribution of Are values for a single galaxy type/ component in each redshift bin. 
The error bars give the limits of the central 68% of data points; the peak of each distribution 
is generally not coincident with the numerical "center" of each bin. Single- component Sersic 
fits to pure bulges and disks recover the radius reliably at all redshifts. Fitted radii to bulge 
and disk components in B+D hosts recover the intrinsic radius of the component to within 1 
kpc on average; however, the errors are larger and these fits are more uncertain. For example, 
the width (containing 68% of sources) of the Are distribution of z ~ 1 pure disks is less than 
1 kpc, whereas the width of the 2; ~ 1 distribution of Are for disk components in B-l-D hosts 

- 30 - 


1 1 1 1 1 1 1 1 1 




1 1 1 1 1 1 1 1 1 1 1 

B-type inactive galaxies 

D-type inactive galaxies 









■ L 





n L 

^-H 1 -1 . — 1 1 — — 1 

1 1 1 1 1 1 1 1 1 



1 1 1 1 1 1 1 1 1 1 1 


20 25 30 35 40 

Fitted m„„ 

Fig. 15. — Distribution of fitted point-source magnitude for simulated inactive galaxies. 
Only 1% of pure disk inactive galaxies (blue dashed histogram) recovers point-source mag- 
nitudes brighter than the lOo" detection limit (dotted line) for the GOODS survey (from 
which our noise properties are taken). When the galaxy is a pure bulge (black histogram), 
however, a point source is found above the lOcr threshold in approximately 25% of fits. The 
overall percentage of fits where a point-source is found when none exists is 12% (for both 
pure bulges and pure disks combined), but this number is strongly dependent on intrinsic 

- 31 - 




1 1 1 1 1 1 1 

1 1 1 1 - 
□ — 




L 1024 AGN hosts 

418 inactive galaxies 

- 1 1 1 1 1 1 1 1 1 1 

1 1 1 1 1 1 1 


0.2 0.4 0.6 0.8 1 

□ Q □ 


1 °- 


639 AGN hosts 

457 inactive galaxies 






redshift z 


Fig. 16. — Dependence of galaxy classification on redshift for simulated bulge galaxies 
(rZin = 4; top, solid lines) and disk galaxies {n-,^ = 1; bottom, dot-dashed lines) with (blue) 
and without (black) central point sources. The y-axis is the fraction of galaxies correctly 
identified as bulges or disks given a threshold of n > 2.0 for bulge-dominated classification 
and n < 2.0 for a galaxy with a significant disk. Although we have here adopted a cutoff 
of n = 2.0, the results change by less than 5% when we change the cutoff to values between 
1.5 < n< 3. 

- 32 - 


"T 1 T 

"T 1 r 

"T 1 T 

"T 1 T 

• Pure Bulges 
B+D Bulges 

"T 1 T 

78.5% _ - - - - ^ 
=^ i 



T 95.3% 







H h 

H 1 h 

H \ h 

H \ h 

H 1 h 

H ^ h 

• Pure Disks 
B+D Disks 

\ 93.6%_ _ - - 




* 90.9% 





-5 - 

_l I L_ 

_l I L_ 


_l I L_ 

_l I L_ 

_l I L_ 



redshift z 



Fig. 17. — Recovery of input radii for pure and B+D bulges (top panel) and disks (bottom 
panel). Each point represents the peak of the distribution of single-component (black, solid 
lines) and two-component (green, dashed lines) Are at each redshift. The numbers next to 
each point indicate the percentage of completed fits with rg > 5 and Are < 90 pixels, so 
that the fit converged to a physical value and would not be rejected by eye based on the fit 
residuals. While median fitted bulge and disk sizes are generally accurate to within 1 kpc at 
all redshifts, the presence of a central point source increases the spread in the distribution 
by up to a factor of 3. In addition, the majority of bulge fits within B+D galaxies at 2; ~ 1 
are not able to converge to a sensible value for the bulge radius. 

- 33 - 

is approximately 3 kpc. Although the average recovery of bulge and disk component radii 
in B+D galaxies is accurate to within 1 kpc, individual radii are highly uncertain. 

The percentage of sources remaining after removing fits with incorrect total magnitudes 
and unphysically low or high values of re, as previously discussed, is also shown for each 
bin in Figure [T71 The number of fits for which the fitted < 5 pixels remains relatively 
constant with redshift, but the number of fits for which the fitted rg value diverges increases 
with redshift, more dramatically so for bulges than for disks. In the z ^ 1 sample, fewer 
than 40% of bulges within two-component fits have fitted radii within a reasonable physical 
range. This is likely due not just to the presence of the additional disk component, but also 
to the fact that the bulge is much smaller at higher redshift compared to the size of the 
central point source. In these cases, the bulge and the point source can be confused, often 
resulting in a divergent bulge fit. This happens less frequently for disks for two reasons: 
first, disks are generally more extended than bulges, and second, the radial profile of the 
ACS PSF more closely resembles that of a de Vaucouleurs bulge than an exponential disk. 

Because different fractions of galaxies are removed from the sample at different redshifts 
(and for different bulge sizes), the morphological composition of a galaxy sample may be al- 
tered from its intrinsic composition simply by applying a reasonable cut on the morphological 
fit parameters, even if this cut is applied evenly (as it is in Figure [T7|) . 

Figure [TS] thus shows the fitted morphological fraction of bulges with redshift of our 
simulated sample of pure bulges and disks for both AGN hosts and inactive galaxies. After 
applying the radius and magnitude cuts above (i.e., removing only unphysical fit results 
from the sample), we calculate the fraction of galaxies in each redshift bin that would be 
classified as bulge-dominated according to an n > 2 cut. For a perfectly-recovered sample, 
this fraction would be N{n > 2.0) /Ntot = 0.5 at all redshifts because we simulated equal 
numbers of bulges and disks. However, the convergence rate for pure bulge and disk fits 
is slightly different at each redshift, and even our set of reasonable and conservative error 
cuts removes different fractions of bulge and disk galaxies at each redshift. While this does 
affect the morphological composition of the sample at each redshift, its effect is small: the 
morphological fraction of n > 2 galaxies for AGN hosts is typically within 2% of the intrinsic 
fraction at all redshifts. The inactive galaxy morphological fraction is within 5% of the 
intrinsic fraction at all redshifts; both samples are consistent within their uncertainties with 
the intrinsic fraction. 

For the two-component B+D galaxies, the dependence of fitted Sersic index on input 
bulge-to-total ratio is shown in Figure [13 We see a general dependence of fitted n on the 
intrinsic bulge-to-total ratio. The relation changes somewhat as the sample is redshifted, 
but at all redshifts the Sersic index indicates that galaxies may be fitted with a Sersic index 


0.6 - 




"T 1 r 

n 1 r 

"T 1 r 

n 1 r 

3535 AGN Host Galaxies 
1049 Inactive Galaxies 




redshift z 


Fig. 18. — Fraction of galaxies classified as bulges with redshift. Blue circles represent 
simulated AGN hosts, and black triangles represent simulated inactive galaxies (error bars 
determined using Poisson statistics). Because the initial simulation contains equal numbers 
of pure bulges and disks, the actual bulge fraction is 0.5 for both samples. At the lowest 
redshift bin, these fractions are recovered to within 1% for both samples. The sample of 
simulated AGN host galaxies is mis- classified by up to 2% (0.02), and the inactive sample 
is mis-classified by up to 5%. Uncertainties are higher for inactive galaxies due to lower 
numbers of galaxies in the inactive sample, but in both cases the recovered fractions are 
consistent with the correct fraction within the uncertainties. 

- 35 - 

z = 0. 125 


z = 0.4125 

z = 0.7375 

4 • 


z= 1.075 

^ i • * 

\ I I I \ L 




0.4 0.6 0. 

input (B/Tot) 

Fig. 19. — Fitted Sersic index vs. input bulge-to-total ratio, for simulated normal galaxies 
(black triangles) and galaxies with added point sources (blue circles). At all redshifts (low 
to high, top to bottom) the fit correctly finds disks {n = 1) at (LB/Ltot) = and bulges 
(n = 4) at (LB/Ltot) = 1- However, some galaxies fit to Sersic values of > 4 despite 
having up to 30% of their intrinsic light from a disk. Galaxies with intermediate values of 
77, (2 < n < 4) likely have bulges that contribute between 20% and 70% of the total galaxy 
light. Points plotted are the median of each bin's distribution; error bars mark the widths 
of each distribution enclosing the central 68% of sources in the bin. 

- 36 - 

indicative of a pure bulge [n = 4) even in cases where a disk is present and contributing up to 
45% of the total galaxy light. However, a fitted Sersic index consistent with an exponential 
disk [n = 1) typically indicates very little bulge contribution (< 10%); galaxies fitted with 
n < 1.5 have {Ls/Ltot) < 0.2. Intermediate values of the Sersic parameter, 1.5 < n < 3, 
generally indicate a galaxy with both bulge and disk, but the bulge-to-total ratio may vary 
between 0.2 < (L^/i^toi) < 0.65, indicating that it is impossible to determine with a single- 
component fit whether a host galaxy with an intermediate Sersic index is intrinsically bulge- 
or disk-dominated. Given that two-component B+D fits are also uncertain in the presence of 
a central point source, we conclude that determination of the bulge-to-total ratio is uncertain 
by at least 20% in AGN host galaxies, with a higher uncertainty for hosts with intermediate 
Sersic indices and/or fitted {Ls/Ltot)- 

Figure [20] shows the recovery of detected input point sources, hosts, and host-to-PS 
contrast ratios for our simulated galaxies (B, D, and B+D) at all redshifts. Again we see 
that for the local sample, the recovery is excellent: not only are point sources detected at 
least 95% of the time, but the recovered magnitude is typically within at least 0.1 ± 0.12 
dex of the intrinsic magnitude. In fact, point-source recovery is very good at all redshifts: 
the recovered magnitude is within 0.5 dex of the input magnitude for all redshifts and input 
point-source magnitudes. 

The dispersion in recovered values for the host galaxy magnitude is higher than that of 
point-source magnitude values for all redshifts. While in the local redshift bin the median 
recovered host magnitude is within 0.1 dex of the input magnitude, the central 68% of 
values has a larger spread. At higher redshifts, the host galaxy is more likely to converge 
to magnitudes that are too bright compared to the input magnitude. This effect is more 
pronounced for sources with a brighter input point source, as expected. For sources with a 
bright point source compared to the host, the difference is Arrihost = 0.6 for z > 0.7. For 
z ^ 1, the distribution of Arrihost has an extended tail to negative (too-bright) values: a 
significant fraction of high- redshift sources overestimates the host brightness by up to 1.8 
dex regardless of the host-to-PS contrast ratio. For z < 1, this tail also exists for all host- 
to-PS values. However, when the input point source is faint compared to the host, the host 
magnitude is fitted to within —0.8 < Arrihost < 0.3 (la) for these redshifts. 

5. Summary and Conclusions 

We simulated 54,418 GOODS ACS galaxy images of inactive and AGN host galaxies 
at redshifts 0.125 < z < 1.25, the largest sample of simulated two-dimensional galaxy 
morphologies to date. 


1 ' ' ' r 

o z = 0.125 

□ z = 0.412 

A z = 0.738 

* z = 1.075 

H \ h 

H \ h 

H \ h 

H \ h 

H \ h 

H \ h 

H \ h 

H \ h 

H \ h 



3 =^ 


' j 

H \ h 

[ =4 

1^ a 


8 M 








Input nip 

Fig. 20. — Effects of redshift on recovery of input point sources (top), input host magnitudes 
(middle), and host-to-PS contrast ratio, shown for all our simulated host galaxies. Each panel 
shows the fitted quantity versus the input point-source magnitude. Each hne in a panel 
represents a different redshift bin: z = 0.125 (blue, circles), z = 0.4125 (green, squares), 
z = 0.7375 (orange, triangles), and z = 1.075 (red, stars). 

- 38 - 

Using a robust initial guess routine followed by batch-fitting, we performed single- 
component and bulge-disk host galaxy fits on these simulated galaxies, while also extracting 
the central point-source components of the AGN. 

For batch-fit galaxies with central point-source components, typically 60%-70% of the 
fits converge successfully and pass our error cuts. This percentage can be significantly 
increased by follow-up case-by-case fitting, starting from the results of the batch fitting 

We reliably extract central point-source magnitudes for the simulated AGNs using a 
simple xl cut on the recovered fits. Using a combination < 2 and relative effective radius 
error cut of (cr,.^/re) = 0.8 removes 88% of the poor host galaxy fits while excluding 12% of 
good host galaxy fits. The accurate extraction of AGN and host galaxy parameters does not 
depend on the host:AGN contrast ratio (Lj^Qg|./Lpg), as long as it is greater than 1:4. 

Galaxy size is recovered to within Ar^ < 1 kpc for single-component galaxies. For indi- 
vidual bulges and disks within two-component bulge -|- disk host galaxy fits, the distribution 
of recovered radii peaks within Are < 1 kpc at all redshifts, but the large width of the dis- 
tributions indicates that individual fits to bulges and disks within two-component galaxies 
are highly uncertain. 

The average recovered morphological fraction of our sample is within 5% of the true 
fraction, even a.t z ^ 1. Of course, this number should be regarded as a lower limit since 
actual data samples of two-dimensional AGN host galaxy morphologies are likely to be 
significantly smaller than our simulations, and thus suffer from more uncertain statistics. 
We recover the correct morphology 99% of the time for inactive galaxies out to z = 1.25. 

The average bulge-to-total ratios are within 10% of the true ratio at all redshifts, but 
with large scatter (~ 50%). This means that values for individual hosts are unreliable at 
the 50% level, but statistical population averages are more accurate. Th us, uncertainties in 

indiv i dual black hole masses determined using Lhv,i„ e. — Mbh relations (IMcLure fc Dunlop 

2002; Marconi fc Hunt 2003: Ferrarese fc Ford 

20051 ) are higher when using fitted bulge lu- 

minosities of AGN host galaxies than those of inactive galaxies, and may be systematically 
overestimated due to the presence of central point sources. 

Galaxies with Sersic n < 1.5 are generally disk-dominated, with at least 80% of their 
total light coming from a disky component. Galaxies with intermediate 1.5 < n < 3 have 
larger bulge components (occupying from 20% to 65% of the total galaxy light). Galaxies 
that appear to be bulge-dominated, i.e., with n > 3, may derive as little as 45% (depending 
on redshift) of their total light from a bulge component; for n = 4, typically ~ 70% of the 
galaxy light comes from the bulge. 

- 39 - 

In fully simulated AGN host galaxies, the central point source is correctly detected over 
90% of the time, with only weak dependence on the intrinsic contrast ratio between the host 
galaxy and point source or the host morphology. This indicates that a sample of AGN host 
galaxy fits to ACS data with z < 1.25 is at least 90% complete in its detection of central 
point sources. 

However, we also detect spurious point sources in simulated galaxies where no central 
point source is present, for as little as 1% or as high as 25% of the galaxies, depending on 
the morphological composition of the sample. Fits to bulge-dominated galaxies are far more 
likely to detect a point source when none are present than fits to disk-dominated galaxies. 

We have used these simulations to evaluate the robustness of our results for AGN host 
galaxies observed with ACS. This is important for assessing the accuracy of recovered host 
morphologies, as well as the evolution of AGN host morphology with redshift. All of our 
results that can be quantitatively applied to a sample of real AGN host galaxies are also 
presented in tabular format (Tables [21 [3l and|4]). 

We have performed these simulations using the image depths and n oise properties o f 
GOODS, b ut they may also be used for other ACS surveys such as GEMS flRix et al.ll2004l ). 
COSMOS JScoville et al.ll2007h . and AEGIS JPavis et al. l boOTh . The extent to which these 
simulations are directly applicable or provide limits on the accuracy of AGN host galaxy 
morphological fits for other ACS data depends on the relative depth of observations in each 

It is important to note that these results apply to redshifts where the ACS bands are 
observing rest-frame optical data, e.g. z < 1.25. As we move into the era of newer telescopes 
such as the James Webb Space Telescope and observing rest-frame optical wavelengths at 
very high resolution becomes possible for higher redshifts, an understanding of the effects 
of a central point source on high-redshift morp hologies will becom e crucial. Simulations to 
probe these effects have begun on small scales ( Dasyra et al.l 2008 ). but more such work is 

The authors wish to thank C. Conselice, L. Moustakas, P. Natarajan, S. Ravindranath, 
and the entire GOODS team for helpful discussions that improved this work. Thanks espe- 
cially to C. Peng for making GALFIT publicly available and for many enlightening discus- 
sions. We acknowledge support from NASA through grants HST-AR-10689.01-A, HST-GO- 
09425. 13-A and HST-GO-09822.09-A from the Space Telescope Science Institute, which is 
operated by the Association of Universities for Research in Astronomy under NASA contract 
NAS 5-26555. 



Alonso-Herrero, A., Perez-Gonzalez, P. G., Rieke, G. H., Alexander, D. M., Rigby, J. R., 
Papovich, C, Donley, J. L., & Rigopoulou, D. 2008, ApJ, 677, 127 

Ballo, L., Cristiani, S., Fasano, G., Fontanot, F., Monaco, P., Nonino, M., Pignatelli, E., 
Tozzi, P., Vanzella, E., Fontana, A., Giallongo, E., Grazian, A., & Danese, L. 2007, 
ApJ, 667, 97 

Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393 

Binney, J., & Merrifield, M. 1998, Galactic astronomy (Galactic astronomy / James Binney 
and Michael Merrifield. Princeton, NJ : Princeton University Press, 1998. (Princeton 
series in astrophysics) QB857 .B522 1998 ($35.00)) 

Conselice, C. J. 2003, ApJS, 147, 1 

Dasyra, K. M., Yan, L., Helou, G., Surace, J., Sajina, A., & Colbert, J. 2008, ArXiv e-prints, 

Davis, M., Guhathakurta, P., Konidaris, N. P., Newman, J. A., Ashby, M. L. N., Biggs, 
A. D., Barmby, P., Bundy, K., Chapman, S. C, Coil, A. L., Conselice, C. J., Cooper, 
M. C, Croton, D. J., Eisenhardt, P. R. M., EUis, R. S., Faber, S. M., Fang, T., Fazio, 
G. G., Georgakakis, A., Gerke, B. F., Goss, W. M., Gwyn, S., Harker, J., Hopkins, 
A. M., Huang, J.-S., Ivison, R. J., Kassin, S. A., Kirby, E. N., Koekemoer, A. M., 
Koo, D. C, Laird, E. S., Le Floc'h, E., Lin, L., Lotz, J. M., Marshall, P. J., Martin, 
D. C, Metevier, A. J., Moustakas, L. A., Nandra, K., Noeske, K. G., Papovich, C, 
Phillips, A. C, Rich, R. M., Rieke, G. H., Rigopoulou, D., Salim, S., Schiminovich, 
D., Simard, L., Small, I., Small, T. A., Weiner, B. J., Willmer, C. N. A., Willner, 
S. P., Wilson, G., Wright, E. L., & Yan, R. 2007, ApJ, 660, LI 

Ferrarese, L., & Ford, H. 2005, Space Science Reviews, 116, 523 

Ferrarese, L., & Merritt, D. 2000, ApJ, 539, L9 

Gebhardt, K., Bender, R., Bower, G., Dressier, A., Faber, S. M., Filippenko, A. V., Green, 
R., Grillmair, C, Ho, L. C, Kormendy, J., Lauer, T. R., Magorrian, J., Pinkney, J., 
Richstone, D., & Tremaine, S. 2000, ApJ, 539, L13 

Giavalisco, M., Ferguson, H. C, Koekemoer, A. M., Dickinson, M., Alexander, D. M., 
Bauer, F. E., Bergeron, J., Biagetti, C, Brandt, W. N., Casertano, S., Cesarsky, C, 
Chatzichristou, E., Conselice, C, Cristiani, S., Da Costa, L., Dahlen, T., de Mello, 

-41 - 

D., Eisenhardt, P., Erben, T., Fall, S. M., Fassnacht, C, Fosbury, R., Fruchter, A., 
Gardner, J. P., Grogin, N., Hook, R. N., Hornschemeier, A. E., Idzi, R., Jogee, S., 
Kretchmer, C, Laidler, V., Lee, K. S., Livio, M., Lucas, R., Madau, P., Mobasher, B., 
Moustakas, L. A., Nonino, M., Padovani, P., Papovich, C., Park, Y., Ravindranath, 
S., Renzini, A., Richardson, M., Riess, A., Rosati, P., Scliirmer, M., Schreier, E., 
Somerville, R. S., Spinrad, H., Stern, D., Stiavelli, M., Strolger, L., Urry, C. M., 
Vandame, B., Williams, R., & Wolf, C. 2004, ApJ, 600, L93 

Graham, A. W. 2001, AJ, 121, 820 

Grogin, N. A., Conselice, C. J., Chatzichristou, E., Alexander, D. M., Bauer, F. E., Horn- 
schemeier, A. E., Jogee, S., Koekemoer, A. M., Laidler, V. G., Livio, M., Lucas, R. A., 
Paolillo, M., Ravindranath, S., Schreier, E. J., Simmons, B. D., & Urry, C. M. 2005, 
ApJ, 627, L97 

Haussler, B., Mcintosh, D. H., Harden, M., Bell, E. F., Rix, H.-W., Borch, A., Beckwith, 
S. V. W., Caldwell, J. A. R., Heymans, C., Jahnke, K., Jogee, S., Koposov, S. E., 
Meisenheimer, K., Sanchez, S. F., Somerville, R. S., Wisotzki, L., & Wolf, C. 2007, 
ApJS, 172, 615 

Jogee, S., Barazza, F. D., Rix, H.-W., Shlosman, I., Harden, M., Wolf, C., Davies, J., Heyer, 
I., Beckwith, S. V. W., Bell, E. F., Borch, A., Caldwell, J. A. R., Conselice, C. J., 
Dahlen, T., Haussler, B., Heymans, C, Jahnke, K., Knapen, J. H., Laine, S., Lubell, 
G. M., Mobasher, B., Mcintosh, D. H., Meisenheimer, K., Peng, C. Y., Ravindranath, 
S., Sanchez, S. F., Somerville, R. S., & Wisotzki, L. 2004, ApJ, 615, L105 

Koekemoer, A. M., Fruchter, A. S., Hook, R. N., & Hack, W. 2002, in The 2002 HST 
Calibration Workshop : Hubble after the Installation of the ACS and the NICMOS 
Cooling System, Proceedings of a Workshop held at the Space Telescope Science 
Institute, Baltimore, Maryland, October 17 and 18, 2002. Edited by Santiago Arribas, 
Anton Koekemoer, and Brad Whitmore. Baltimore, MD: Space Telescope Science 
Institute, 2002., p.339, 339-+ 

Kormendy, J., & Richstone, D. 1995, ARA&A, 33, 581 

Magorrian, J., Tremaine, S., Richstone, D., Bender, R., Bower, G., Dressier, A., Faber, 
S. M., Gebhardt, K., Green, R., Grillmair, C, Kormendy, J., & Lauer, T. 1998, AJ, 
115, 2285 

Marconi, A., & Hunt, L. K. 2003, ApJ, 589, L21 
Marleau, F. R., & Simard, L. 1998, ApJ, 507, 585 


McLure, R. J., & Dunlop, J. S. 2002, MNRAS, 331, 795 

Peng, C. Y., Ho, L. C, Impey, C. D., & Rix, H.-W. 2002, AJ, 124, 266 

Ravindranath, S., Ferguson, H. C, Conselice, C, Giavalisco, M., Dickinson, M., Chatzichris- 
tou, E., de Mello, D., Fall, S. M., Gardner, J. P., Grogin, N. A., Hornschemeier, A., 
Jogee, S., Koekemoer, A., Kretchmer, C., Livio, M., Mobasher, B., & Somerville, R. 
2004, ApJ, 604, L9 

Rix, H.-W., Harden, M., Beckwith, S. V. W., Bell, E. F., Borch, A., Caldwell, J. A. R., 
Haussler, B., Jahnke, K., Jogee, S., Mcintosh, D. H., Meisenheimer, K., Peng, C. Y., 
Sanchez, S. F., Somerville, R. S., Wisotzki, L., & Wolf, C. 2004, ApJS, 152, 163 

Sanchez, S. F., Jahnke, K., Wisotzki, L., Mcintosh, D. H., Bell, E. F., Barden, M., Beckwith, 
S. V. W., Borch, A., Caldwell, J. A. R., Haussler, B., Jogee, S., Meisenheimer, K., 
Peng, C. Y., Rix, H.-W., Somerville, R. S., & Wolf, C. 2004, ApJ, 614, 586 

Scoville, N., Aussel, H., Brusa, M., Capak, P., Carollo, C. M., Elvis, M., Giavalisco, M., 
Guzzo, L., Hasinger, G., Impey, C, Kneib, J. -P., LeFevre, O., Lilly, S. J., Mobasher, 
B., Renzini, A., Rich, R. M., Sanders, D. B., Schinnerer, E., Schminovich, D., Shop- 
bell, P., Taniguchi, Y., & Tyson, N. D. 2007, ApJS, 172, 1 

Sersic, J. L. 1968, Atlas de galaxias australes (Cordoba, Argentina: Observatorio Astro- 
nomico, 1968) 

Sheth, K., Elmegreen, D. M., Elmegreen, B. G., Capak, P., Abraham, R. G., Athanassoula, 
E., Ellis, R. S., Mobasher, B., Salvato, M., Schinnerer, E., Scoville, N. Z., Spalsbury, 
L., Strubbe, L., Carollo, M., Rich, M., & West, A. A. 2008, ApJ, 675, 1141 

Simard, L., Koo, D. C, Faber, S. M., Sarajedini, V. L., Vogt, N. P., Phillips, A. C, Gebhardt, 
K., lUingworth, G. D., & Wu, K. L. 1999, ApJ, 519, 563 

Spergel, D. N., Verde, L., Peiris, H. V., Komatsu, E., Nolta, M. R., Bennett, C. L., Halpern, 
M., Hinshaw, G., Jarosik, N., Kogut, A., Limon, M., Meyer, S. S., Page, L., Tucker, 
G. S., Weiland, J. L., WoUack, E., & Wright, E. L. 2003, ApJS, 148, 175 

Trujillo, I., & Aguerri, J. A. L. 2004, MNRAS, 355, 82 

This preprint was prepared with the AAS I^TJt;X macros v5.2. 


Pure Bulges 

Pure Disks 



-f Disk 



16.898, 5.0 

16.898, 5.0 

16.898, 5.0 

16.898, 5.0 

(J T \ 

[■lAB, J^*) 

18.645, 1.0 

18.645, 1.0 

18.645, 1.0 

18.645, 1.0 

19.398, 0.5 

19.398, 0.5 

19.398, 0.5 

19.398, 0.5 

20.393, 0.2 

on ono f\ o 

20.393, 0.2 

on ono n o 

20.393, 0.2 

c\r\ 'T V r\ r\ A 

20.750, 0.14 

on ono n o 

20.393, 0.2 
20.750, 0.14 

13.54, 1.5 

36.12, 4.0 

0*7 r\f\ o r\ 

27.09, 3.0 

r /I 1 o n 

54.18, 6.0 

/•II N 

(pixels, kpc) 

27.09, 3.0 

54.18, 6.0 

40.63, 4.5 

72.23, 8.0 

40.63, 4.5 

72.23, 8.0 

54.18, 6.0 

90.29, 10.0 

54.18, 6.0 

90.09, 10.0 

7 / 







































Table 1: Grid values for simulated galaxies at z = 0.125. The effective radius values differ 
between the bulge and disk galaxies due to the different physical sizes of these two classes of 
galaxies. The parameter space varies slightly between single-component galaxies and double- 
component galaxies in order to increase the number of mesh points in the bulge-to-total ratio 
without increasing the simulated double- component galaxies to a computationally prohibitive 
number. Conversion from magnitude to luminosity and between pixels and physical size uses 
a concordance cosmology {fltot = 1, = 0.73, and Hq = 71 km/s/Mpc) and a redshift 
z = 0.125. 


OUUiL-C IlipQ 


' con?; 




i V2 









































































Table 2: RMS values for fit parameters vary with central point-source magnitude. N^onv is 
the number of fits which converged (out of 450) for each input point-source magnitude value. 
A^i is the number remaining after applying a < 2.0 cut; N2 is the number remaining after 
further requiring ar^/r^ < 0.8. The RMS value for the point-source magnitude is calculated 
from A''i galaxies, and the values for (Jmhost^ and are calculated from N2 objects. Note: 

Resu lts for these magnitudes are based on a survey with the GOODS depths (iGiavalisco et al. 
2004h . 

Source log (L^ost/^Ps) 







































Table 3: Determined RMS values for fit parameters to sources with different log (-^^^ost/ -^PS)- 


Fitted Sersic index 

Intrinsic bulge-to-total ratio 

n < 1.5 
1.5 < n < 3 
n > 3 
n > 4 

(B/Tot) < 0.2 
0.2 < (B/Tot) < 0.65 
(B/Tot) > 0.45 
(B/Tot) > 0.55 

Table 4: Relationship of a host galaxy's fitted Sersic index to its intrinsic bulge-to-total light 
ratio. Hosts with intermediate Sersic index may have a wide range of bulge-to-total ratios; 
hosts with fitted n = 4, typically classified as pure deVaucouleur bulges, may in fact have 
bulge-to-total ratios as low as 55%.